The energy density scatter plot is physically correct but misleading and everyone makes this mistake.

From an engineering point of view you have to use work delivered at the end of the drive train not fuel raw energy content.

When you do that lithium ion batteries compare more favorably to liquid fuels. That’s because the conversion path is more than 90% efficient. For ICE engines you’re starting with only 20-40% Carnot efficiency (depending on how good and in good shape the engine is) and then losing in the transmission and then losing more because ICE cars have more other gears and moving parts. Power to wheel is pretty terrible. Most of the energy from gasoline heats the air around the car.

This is also why you get outrageous sounding but accurate things like: an EV charged on 100% coal fired electricity emits less carbon than a typical gasoline car. The fact that coal is literal pure carbon fuel is made up for by the high thermal efficiency of a giant supercritical steam turbine vs a small piston engine. Coal burns real hot too (steeper thermal gradient). So more of the energy from coal ends up doing actual work vs heating the air. (Well directly heating the air I mean.)

The flavour you want is polymer, single ion conducting solid state with an ion transport activation energy below 10kJ/mol at room temperature and no phase transitions from -40C to 80C.

That is the holy grail of SS batteries.

is LiFePo4 immune to dendrite shorts? Or do they happen just don't burn?

Ambri was working on large scale batteries which seemed like a pretty good idea (looks like they ran out of money): https://en.wikipedia.org/wiki/Ambri_Inc.